Your question is Feature Engineering for Sparse Data. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
You're working on a supervised learning task where most features come from high-cardinality categorical fields and text-like signals, so the design matrix is very high-dimensional and mostly zeros. You need a practical way to create useful features without overfitting or making training too expensive.
How do you approach feature engineering for high-dimensional, sparse datasets?